Note-and-vote / Decider supervote

Status: Documented, not shipped · Evidence: P · Family: Facilitation and group structures · Verdict: reject (2026-06-09)

What it is

Note-and-vote is a group-facilitation protocol from the Google Ventures (GV) design sprint, popularized by Jake Knapp in Sprint (2016) and on GV’s Design Sprint Kit. It is a fixed sequence a facilitator runs to get a roomful of people from many options to one decision without the meeting collapsing into the usual failure modes (the loudest voice winning, endless circular debate, premature consensus). The canonical steps are: each person silently and independently writes their own notes or candidate solutions; each person silently narrows to a personal shortlist; everyone places votes - often dot votes - on the options laid out on the wall; each voter explains their vote in roughly one minute while the others listen; and then a single designated Decider casts a weighted “supervote” (in the GV formulation, three marked votes the Decider may stack on one option or spread across several) that settles what the team actually pursues. The Decider may respect the group’s votes or override them; the explicit doctrine is that “you need a decision-maker in the room” and the group’s votes are advice to the Decider, not a binding tally.

The durable claim of the method is sequencing and authority, not idea content. By forcing silent individual work before anyone speaks, it tries to defeat anchoring on the first or most senior speaker; by collecting votes before open discussion, it tries to surface the spread of genuine private preference rather than a manufactured consensus; and by ending on a single accountable Decider rather than a majority count, it tries to avoid the design-by-committee average. Every one of those payoffs is a property of a group of independent people being run through the structure. The artifact the method emits is a wall of dotted options and a Decider’s final mark - a recorded group choice - not an analytical deliverable a solo reasoner produces.

When it helps / when it misleads

As a meeting structure for a real human group, note-and-vote helps when several people hold relevant private information or preference, the options are already on the table (it is a converge move, not a generate-from-blank move), and the meeting is at risk of being dominated by hierarchy or by the first articulate opinion. The silent-first, vote-before-debate ordering is a credible guard against the HIPPO effect (the Highest-Paid Person’s Opinion) and against the conformity cascade, and the single-Decider ending gives a fast, accountable close where a consensus-seeking group would stall or average.

It misleads or simply does not apply when:

• There is no group. The entire mechanism is the management of many independent humans’ inputs and one human’s authority. A single reasoner (or a single AI agent) has no other voters to de-anchor from, no private-information spread to pool, and no Decider role to confer. Run “solo,” the protocol degenerates into “list some ideas, then pick one,” which is not what the method is for.
• The options have not been generated yet. Note-and-vote converges; it presumes sketches or candidates already exist (in the sprint, it follows the sketching day). Pointed at a blank page it does nothing.
• The vote is mistaken for the decision. The method’s own doctrine is that the dots are advice and the Decider decides; teams that treat the dot count as binding have quietly swapped in plurality voting, which the evidence (below) finds worse than the multi-vote spread the method actually prescribes.
• The structure substitutes for diverse inputs. If everyone in the room already shares the same information and frame, no voting ritual recovers a hidden better option; the protocol surfaces preferences, it does not manufacture the unshared knowledge a hidden-profile situation needs.
• Dot voting’s own biases go unmanaged. Visible dots invite herding and late-voter bandwagon effects unless votes are cast simultaneously or concealed - a known weakness of open dot voting that the basic ritual does not fix.

What the evidence says

The honest grade for note-and-vote’s own move - “run a group through silent notes, then a vote, then a Decider supervote” - is P (practitioner), and the grading has to be careful, because the method’s popular write-ups borrow robustness from neighboring research that measured different operations or located the benefit in a different place than the part note-and-vote adds.

What the record supports. Note-and-vote is a real, named, widely-taught facilitation protocol with a clear origin (GV / Knapp). Its silent-individual-generation-before-discussion step rests on the better-evidenced Nominal Group Technique tradition (Delbecq and Van de Ven, 1971), where requiring people to write independently before the group talks reliably raises the number and diversity of ideas by blunting conformity - but that step is exactly the brainwriting / NGT mechanism the library already ships, not the part note-and-vote uniquely contributes. On the voting half, the strongest controlled evidence is Johnson, Awtrey and Ong (2022) in Academy of Management Discoveries: across 93 student groups simulating counterterrorism “pursuit teams,” groups that took an unofficial multivote (ten votes spread across three options) were about 50% more likely to identify the correct option than groups using plurality or ranked-choice voting. Crucially for note-and-vote’s distinctness claim, the authors report the gain appeared before any discussion - allocating multiple votes made individuals process the evidence more deeply on their own, rather than the group’s deliberation doing the work. That finding endorses multi-voting over plurality (note-and-vote does use a spread vote, so this is in its favor) but it locates the value in a solo deep-processing effect, and it says nothing about the Decider-supervote authority step.

What the record does NOT support, and the laundering trap. There is no controlled or comparative study I can locate that measures note-and-vote as a protocol (silent notes -> vote -> Decider supervote) against an alternative meeting, and none that measures the Decider supervote - the one element that distinguishes it from plain NGT-plus-multivote - against majority rule or consensus. The widely-circulated “design sprints save ~7x the time” figure traces to a self-reported retrospective survey by the agency AJ&Smart (participants estimating how long the work “would have” taken), not to any primary experiment; it has no controlled baseline and is excluded from the grade. The genuinely experimental results that do exist measure adjacent operations and must not be transferred onto note-and-vote’s own move: the multivote result above is about voting method, not the Decider step; the hidden-profile literature (Stasser and Titus, 1985; meta-analyzed by Lu, Yuan and McLeod, 2012, finding groups mention far more shared than unshared information and are several times less likely to find the hidden-profile solution) explains why a private-vote-first structure could help, but it tests information pooling, not this protocol, and it equally implies that a vote alone cannot rescue a group that never surfaces the unshared facts. Borrowing any of those grades to lift note-and-vote above P would be laundering a cousin’s robustness onto a move the cousin did not test. The conservative governing grade is therefore P: a recognized practitioner protocol, with one supportive multivote study that actually points at a solo effect, and no direct controlled evidence for the protocol or its signature Decider step.

Transfer caveat (required). Every study above is on human groups in lab or field settings (students, juries, intelligence-team simulations). None studies note-and-vote, multivoting, or a Decider supervote performed by or with an AI agent, and the method’s whole value rests on social dynamics among independent humans. The evidence is transferred from human group contexts and is not validated - and largely not even applicable - to a single AI agent.

Excluded figures. The “~7x time savings” (AJ&Smart self-reported survey) and the looser “2x/4x/6x ROI” figures from vendor design-sprint studios have no primary experimental source and are excluded; they may not influence the grade.

Why it is / is not a skill here

Verdict: Reject (status excl). This overturns the catalog’s prior cand / build / P tag, whose own one-line rationale - “Candidate; an agent can support capture” - already conceded the problem: the only role the note offered an agent was stenographer for a human group’s vote, which is not a thinking move the agent performs.

The Build burden is to name one distinct, durable cognitive move that a solo agent performs and that no shipped skill already produces, plus the separable artifact it emits. Note-and-vote fails that burden on both the artifact test and the agent-applicability test, and it splits cleanly into two halves that are each already disposed of:

• The generation half is already owned by brainwriting. “Each person silently writes candidates before anyone speaks” is the Nominal Group Technique / silent-writing-before-discussion mechanism, and the registry has already absorbed both of those into brainwriting (whose entry name is literally “Brainwriting 6-3-5 / NGT” and whose aliases include “NGT” and “Nominal Group Technique”; silent-writing-before-discussion is a recorded fold into it). So the capturable, agent-runnable part of note-and-vote adds no new move - it is brainwriting.

• The distinctive half is a group social-governance ritual the catalog has already ruled non-agent-reproducible. What note-and-vote adds over brainwriting is the dot-vote aggregation and the Decider supervote. But dot-voting is already flag / reject here (“herding caveats”), and the family’s own exclusion of scaled-participation-formats states the governing principle outright: such formats are excluded because “the value is human social dynamics an AI cannot reproduce.” The Decider supervote is the purest case of that - its entire function is conferring decision authority on one human and aggregating other humans’ votes, neither of which exists for a single agent. There is no separable solo artifact: strip the group away and the protocol is “generate options, then choose one against preference,” which is brainwriting followed by decision-option-review - a recipe of two shipped moves, whose irreducible remainder (the anti-HIPPO, anti-groupthink, single-accountable-Decider governance) is exactly the human-dynamics part the library does not ship.

Why reject (excl) rather than fold: a fold into brainwriting would falsely assert that brainwriting captures note-and-vote, when the vote-and-Decider step is the whole point of the method and is precisely the part that is excluded as non-reproducible - folding would hide the exclusion rather than record it. And it is not a clean single-skill subsumption (the generation half lands on brainwriting, the converge half on dot-voting/decision-option-review, the remainder on the excluded social formats), so naming one foldInto target would misrepresent it. Reject is the more honest disposition, mirroring scaled-participation-formats: a real, useful meeting structure that is out of scope for a library of solo-agent thinking artifacts. The learning value of the NO: a famous, genuinely effective group-facilitation protocol is not automatically an agent skill. Note-and-vote’s effectiveness lives in managing many independent humans and one accountable decider; an AI agent has neither, so the library documents the method, points its capturable half at brainwriting and its converge half at decision-option-review, and excludes it rather than ship a hollowed-out ritual under a famous name.

Lineage and who to read

The protocol originates with Jake Knapp and the Google Ventures design team and is set out in Sprint: How to Solve Big Problems and Test New Ideas in Just Five Days (Knapp with John Zeratsky and Braden Kowitz, 2016) and on GV’s Design Sprint Kit (“Note and Vote,” Decide phase); the popular write-up that named it for a wide audience is the Fast Company piece “‘Note And Vote’: How Google Ventures Avoids Groupthink In Meetings.” For the better-evidenced ancestor of its silent-generation step, read Andre Delbecq and Andrew Van de Ven on the Nominal Group Technique (1971) - the same lineage the shipped brainwriting skill cites. For the voting half, read Michael D. Johnson, Eli Awtrey and Wei Jee Ong, “Verdicts, Elections, and Counterterrorism: When Groups Take Unofficial Votes” (Academy of Management Discoveries, 2022), the multivote study whose gain appeared before discussion. For why a vote-before-debate structure can help and where it cannot, read Garold Stasser and William Titus on the hidden-profile / shared-information bias (1985) and the 25-year meta-analytic review by Li Lu, Y. Connie Yuan and Poppy McLeod (2012). “Note-and-vote,” “Decider,” and “supervote” are descriptive design-sprint terms in common use, attributed to Knapp / GV; there is no registered trademark to clear, so this entry is documented descriptively and is not flagged as branded.

Named sources

• Jake Knapp, John Zeratsky and Braden Kowitz, Sprint: How to Solve Big Problems and Test New Ideas in Just Five Days (Simon and Schuster, 2016). The primary articulation of note-and-vote and the Decider supervote as a sprint convergence ritual. Practitioner / foundational. (P)
• Google Ventures, “Note and Vote” (Design Sprint Kit, Decide phase), designsprintkit.withgoogle.com. The canonical step-by-step of the protocol (silent notes -> shortlist -> vote -> explain -> Decider supervote). Practitioner / primary doctrine. (P)
• Belle Beth Cooper / Fast Company, “‘Note And Vote’: How Google Ventures Avoids Groupthink In Meetings” (2014), fastcompany.com. The popular write-up that defined the term; describes votes as advice to the Decider, who decides. Practitioner / popular. (P)
• Andre L. Delbecq and Andrew H. Van de Ven, “A Group Process Model for Problem Identification and Program Planning,” Journal of Applied Behavioral Science 7(4) (1971): 466-492. Origin of the Nominal Group Technique: silent independent writing before discussion raises idea quantity and diversity by reducing conformity. The evidenced ancestor of note-and-vote’s generation step - but it is the mechanism brainwriting already absorbs, not the part note-and-vote uniquely adds. (M, for NGT - not for the Decider step)
• Michael D. Johnson, Eli Awtrey and Wei Jee Ong, “Verdicts, Elections, and Counterterrorism: When Groups Take Unofficial Votes,” Academy of Management Discoveries (2022). Experimental: across 93 student “pursuit teams,” multivoting groups were ~50% more likely to pick the correct option than plurality or ranked-choice groups, and the benefit appeared before discussion (deeper individual processing). Supports multi-voting over plurality, but locates the gain in a solo effect and does not test the Decider supervote. (M, for the voting-method effect on human groups - not for note-and-vote’s signature step)
• Garold Stasser and William Titus, “Pooling of Unshared Information in Group Decision Making: Biased Information Sampling During Discussion,” Journal of Personality and Social Psychology 48(6) (1985): 1467-1478. Foundational hidden-profile result: groups over-discuss shared information and frequently fail to surface the unshared facts that identify the best option. Explains the rationale for a private-vote-first structure; tests information pooling, not this protocol. (M)
• Li Lu, Y. Connie Yuan and Poppy Lauretta McLeod, “Twenty-Five Years of Hidden Profiles in Group Decision Making: A Meta-Analysis,” Personality and Social Psychology Review 16(1) (2012): 54-75. Meta-analysis confirming the shared-information bias is large and robust across studies. Cited to locate where the “structured group voting helps” intuition is actually evidenced - the information-pooling literature, on human groups. (M)

Excluded under the evidence rule: the popular “design sprints save ~7x time” claim (AJ&Smart self-reported retrospective survey) and the looser “2x/4x/6x ROI” figures from design-sprint vendors have no primary experimental source measuring note-and-vote and are not counted toward this entry’s grade.